We have previously highlighted many potential issues with artificial intelligence (AI), some of which are related to AI translation.
It is already known that Large Language Models (LLMs) can produce inaccurate information, a tendency that is referred to as “hallucinating”. It’s easy to see how fake news could be created and then proliferate.
We could find ourselves living in a world in which it is almost impossible to separate fact from fiction.
On a more mundane level, it is inevitable that AI will increasingly be used to translate content. We know that the technology can be less than accurate. In addition, LLMs certainly cannot yet rival humans in their understanding of both context and the nuances of languages.
Translations that haven’t benefited from human post-editing could prove to be anything from amusingly clunky to wholly inaccurate.
While there may be trouble ahead with AI, not least for translators whose work will disappear, there are aspects of translation where AI could be a major boon.
Indeed, the technology could solve problems that it would be impractical and too expensive to address with human translators.
Translating in real time
When online content, documents, instructions manuals and subtitles are translated, outputs can be checked and edited before the material is published.
What about situations that require translations to be undertaken in real time?
During significant events such as conferences at the United Nations or debates in the European Union, human interpreters are always on hand to translate what is said in real time.
But in everyday situations, things are very different. For instance, It simply wouldn’t be possible to position interpreters at railway ticket offices to assist foreign visitors. Imagine how many interpreters you would need to deal with even the most widely spoken languages!
Here’s where AI could come into its own.
Tokyo AI experiment
A face to face translation tool is being tested in a Tokyo train station.
The device has been installed on the counter of a ticket booth at Tokyo’s Seibu-Shinjuku Station. AI technology is being utilised to enable travellers to converse with an agent while speaking in their own languages. The agent’s responses are translated and then displayed on a transparent screen.
Travellers select their language on an adjacent tablet. The tech can translate a dozen languages including English, Chinese, Korean, Thai, Filipino and French. Both the tourists and the agent can view responses on the screen. Translated responses appear as speech bubbles in each participant’s field of vision. The system facilitates a fluid and natural dialogue.
Better still, the device can also display responses in Japanese, a great benefit for hearing-impaired locals.
The new system will be trialled for a few months and if the trial proves to be successful, the system will be installed at further stations across Japan.
Translations for hearing impaired travellers
Travelling can be stressful for anyone as transport systems are prone to disruption. Everything from cancelled trains to signal failures can interrupt journeys and cause confusion.
Most of us manage to deal with such problems with relative ease. We listen to the announcements and then act accordingly.
We might be filled with dread when we hear the annoying bing-bong sound that tends to precede announcements at airports and railway stations. But when we are told that our departure gate has been changed, we simply join the hoards drifting across the concourse.
But what if we can’t hear the announcements?
The hard of hearing can really struggle when things start to go wrong at railway stations and airports. These situations can be frightening and even dangerous for them. Written information is sometimes available, but for many deaf people, sign language is their first language, not the written language that is displayed on boards and screens.
Network Rail in the UK is one of several organisations around the world now attempting to address this accessibility issue using AI.
The company has started to roll out travel advice screens that deliver messages in British Sign Language (BSL). These screens have been installed at eight stations and they display both standard announcements and information regarding disruptions.
Signapse, a British tech startup is one of the companies behind the new technology being used. It is working to develop effective, real-time sign language translation software in partnership with organisations for the deaf. The synthetic signers that the company has evolved are convincingly human, unlike previous signing avatars that are not favoured by the deaf community.
Video messages for the system are pre-recorded but AI is used to improve the fluency of the recordings. As the AI technology is developed, real-time translation should become possible.
Over 70 million people across the world are deaf and there are up to 300 different sign languages used by those people. The potential for AI translations to create greater accessibility is clear. Work continues into how the capability of AI can be used to develop effective software for translating and generating sign languages.
Translating sign languages presents developers with far greater challenges than tackling text or the spoken word. Many difficulties must be overcome but AI could be the key to a better future for deaf travellers around the world.
Both dangerous and helpful
Artificial intelligence is evolving at a truly remarkable pace. The technology cannot yet deliver translations that rival those of skilled human professionals, and AI could prove to be very dangerous indeed. However, it could also facilitate the development of many incredibly helpful systems that it would not be practical or even possible to offer without the assistance of the technology.
Related Posts
